midreal’s-gen-ai-‘choose-your-own-adventure’-platform-launches

MidReal’s Gen AI ‘choose Your Own Adventure’ Platform Launches

Join gaming leaders live this May 20-21 in Los Angeles to examine the strategies needed to adapt and excel in an ever evolving landscape, featuring insights from leading voices and thought leaders in the industry. Register here.


Generative AI storytelling platform MidReal has officially launched its flagship “choose your own adventure” product. Powered by its latest Morpheus-1 model, the tool allows players to generate stories that evolve as they make decisions. Instead of passively consuming media, MidReal puts gamers, readers, tv and movie buffs, and more in control to shape the narrative.

Founded in October 2023 with researchers from MIT, Cornell and Duke at its disposal, MidReal is blurring the lines between text adventures and games. The San Francisco-based company has seven full-time and five part-time employees to push the potential of the medium.

In addition to story generation, MidReal has more features to boost immersion and keep users engaged. These include generated illustrations with consistent characters and the ability to remix public stories.

“In the world of AI driven narrative and storytelling, readability and believability is of paramount concern and that desire to give creatives the most state-of-the-art tools available is what drove us to create MidReal and its latest AI module ‘Morpheus-1,’” said Kaijie Chen, cofounder and CEO at MidReal. “With its official launch today we’re looking forward to the countless stories that are just waiting to be told to the world.”

GB Event

Countdown to GamesBeat Summit

Secure your spot now and join us in LA for an unforgettable two days experience exploring the theme of resilience and adaptation. Register today to guarantee your seat!

Register Here

MidReal’s academic chops

Unlike other storytelling-focused generative AI tools, the team behind MidReal developed proprietary tools to improve outputs.

“We post-trained large language models to fundamentally transform them from chat-based models into story completion agents optimized for narrative coherence and immersion,” Chen told GamesBeat.

“To do this, we drew inspiration from Shunyu Yao’s work on Tree of Thoughts, which organizes the general problem-solving process into a tree structure. We discovered that the process of story writing is much like constructing a problem-solving tree. In story writing, the user’s input scenario serves as the root of the tree, while subsequent possibilities of plot development serve as the leaf nodes of the tree, spanning into different directions. The process of writing a good story is the process of selecting the branch that has the best plot development.”

MidReal’s Morpheus-1 engine.

Morpheus-1 is a practical application based on a 2023 paper that proposed a post-training framework for creating LLM-based agents. Collaborators on the project included Princeton’s Karthik Narasimhan (one of the original authors of the original GPT study), Shunyu Yao and MidReal’s other cofounder, CTO and President Andrew Chen.

Testing Morpheus-1

True to MidReal’s academic roots, the team tested Morpheus-1’s output against its competitors. In a double-blind study, American readers compared 900 story pairs — one produced by Morpheus and other from a competing model. Readers evaluated these pairs on five dimensions: interestingness, coherence, relevance, emotional impact and language style.

Overall, Morpheus-1 outperformed both GPT-4 and other state of the art storytelling models. Readers preferred MidReal’s product 85% compared to GPT-4 and 75% of the time to competing storytelling products.

“Our model wins the most in the ‘Interestingness’ and ‘Emotional Impact’ direction, which we think are the two most important directions in today’s AI storytelling field,” said Chen. “This is because tools like ChatGPT can indeed tell a story, but the story is usually, if not always, too boring. For a story to actually be consumable, it first needs to be interesting. Hence ‘interestingness’ is the direction that we try to optimize the model the most.”

What’s next

While MidReal is launching today, the San Francisco-based company has no plans of slowing down.

“MidReal is poised to expand its capabilities significantly. We’re working on integrating photo-based fantasy diary composition and interactive video experiences. Additionally, we plan to release our APP and introduce a new ‘SERVER’ feature,” said Chen.

“Imagine entering a Discord server, where your friends in the server are AI agents, and that you can chat with them freely like you do with real-world friends. There will be stories generated in the server, just like social media posts (image and video) and YouTube Shorts. Of course, you can also invite human friends to join the server. This is an immersive world for everyone to live their imagined lives.”

MidReal did raise funding previously and the company is considering a new round of funding in the future.

Try MidReal for free today.

VB Daily

Stay in the know! Get the latest news in your inbox daily

By subscribing, you agree to VentureBeat’s Terms of Service.

Thanks for subscribing. Check out more VB newsletters here.

An error occured.